This page highlights some of the more advanced features of the GramTrans translation software. Unlike word-to-word list-based transfer or statistical translation systems, the GramTrans software uses contextual rules to distinguish between different translations of a given word or phrase.

Core Linguistic Features

• Robust source language analysis.
• Morphological and semantic disambiguation.
• Large linguist-made grammars and lexica.
• High degree of domain-independence: journalistic, literary, email, scientific, etc.
• Name recognition and protection.
• Compound word recognition and seperation.
• Dependency formalism for deep syntactic analysis.
• Context-sensitive selection of translation equivalents.
• Insertion, deletion, and splitting of words to account for language-specific differences.
• Word and phrase reordering for target language output.
• Ability to customize terminology on a per-user basis.

Data Types and User Interfacing

• Plain text.
• Formatted documents and other rich text, such as from MS Word and OpenOffice.org.
• Web pages, including URLs, browser plug-ins, and uploaded HTML documents.
• Movie subtitles, and other custom formats.
• Arbitrary XML structures.
• SMS, WAP, and other mobile protocols.
• Remote API access for integration with external systems.

Example 1

• Danish input: “Står det til Rigspolitiet, må hjemmeværnets 20.000-25.000 medlemmer ikke længere øve sig med løst krudt uden for de militære øvelsesområder.”
• English output: “If it’s up to the State Police, the Home Guard’s 20.000-25.000 members must no longer practise with blank cartridges outside the military practise areas.”

Areas that show use of features:

• “Står det til” → “If it’s up to” shows an insertion of the word “if” to make the text more natural.
• “Rigspolitiet” → “the State Police” shows name recognition and article insertion.
• “må” → “must” shows that words can be moved to account for target language syntax.
• “hjemmeværnets” → “the Home Guard’s” shows recognition of lower-cased names.
• “øvelsesområder” → “practise areas” shows compound recognition and word splitting.

Example 2

• Danish input: “Folkeskolens formålsparagraf skal ændres, fastslår statsminister Anders Fogh Rasmussen efter to dages møde i det såkaldte Globaliseringsråd.”
• English output: “Primary School’s objects clause is to be altered, prime minister Anders Fogh Rasmussen states after two days’ meeting in the so-called Globalisation Council.”

Areas that show use of features:

• “formålsparagraf” → “objects clause” shows compound recognition and word splitting.
• “fastslår” → “states” shows that words can be moved to account for target language syntax.
• “Anders Fogh Rasmussen” → “Anders Fogh Rasmussen” shows name recognition and preservation of multi-word names.
• “Globaliseringsråd” → “Globalisation Council” shows multi word name translation.